为了解决红外图像在图像配准中对比度低、背景复杂、红外目标受噪声干扰严重、传统分割方法易产生过分割或欠分割的问题, 提出了一种基于改进的脉冲耦合神经网络(PCNN)和形态学方法的红外图像分割算法。首先根据图像能量分布情况提取纹理图像, 将纹理图像通过PCNN进行分割, PCNN的链接强度根据区域能量在梯度场的变化自适应设定;由于PCNN的点火位置集中于红外目标部分, 通过点火映射图可以得到连贯清晰的红外目标轮廓;再通过形态学方法滤除背景干扰。结果表明, 该方法能够精确分割红外图像, 分割结果优于传统方法。

Image registration of infrared images have low contrast, complex background and serious noise interference. Over-segmentation or under-segmentation is prone to occur with traditional segmentation method. In order to solve the problems, an improved infrared image segmentation algorithm was proposed based on pulse coupled neural network (PCNN) and morphological methods. Firstly, texture sub-image was extracted according to energy distribution of the image and the texture sub-image was segmented by PCNN. The adaptive links strength of PCNN was set based on the changes of regional energy in gradient field. Because of the firing position of PCNN focused on infrared target portion, a clear coherent infrared target contour can be obtained from firing maps. Background interference was filtered out by morphological methods and high precision infrared target segmentation was achieved. The experimental results show that infrared image can be segmented accurately based on this method. By comparison, the segmentation result is better than traditional methods.